The objective of the study as to assess the consistency with which a set of pre-defined data about three fictitious patients was entered into a sample group of practice computer systems, and to measure the time required for routine data capture of this kind. The study design was a prospective, piloted, postal survey, in which respondents were requested to enter a variety of general sample data onto their systems, to time the process, and record details of any difficulties. The subjects were 76 (39%) responding general practices in England and Ulster, from a random sample of members of a GP computer specialist group. These results (which in view of the highly motivated characteristics of the responding sample are likely to represent best practice) showed that differing conventions were applied in entering patient data, even among practices using the same type of computer system. Potentially significant errors and distortions were found in the data as recorded in the systems, such as the 29% of immunisation sequences which were slightly inaccurately entered. The main problems with the data entry comprised simple operator errors, inconsistency in the use of terms entered, and difficulties in recording negative data (only 20% of practices could enter patient ‘not incontinent’). Practices varied widely in the way they allocated data recording responsibilities to staff, with only 6% of practices involving all staff in data entry. The times reported for entering records varied con siderably, with an average of 26 seconds per data item (ranging from an average 10 seconds per data item to 60 seconds between doctors). The security of computer data was not a particular priority for 76% of practices. In conclusion, the workload of maintaining a fully computerised set of patient records can be substantial. The use of a computerised system for data capture by no means guarantees a standardisation of approach, either within or between practices.